An xDSL is a high-speed data transmission technology over a telephone twisted-pair cable. A central office of an xDSL system, that is, a digital subscriber line access multiplexer (Digital Subscriber Line Access Multiplexer, DSLAM for short), provides access for multiple xDSL signals. Due to an electromagnetic induction principle, mutual interference is generated between the multiple signals, which is referred to as crosstalk. Far-end crosstalk (Far-end Crosstalk, FEXT for short) has relatively large impact on transmission performance of a line. To ensure a channel transmission rate, cancellation processing needs to be performed on the FEXT. Currently, a vectored digital subscriber line (Vectored-DSL) technology is proposed in the industry, which is mainly used to perform joint signal processing at a DSLAM end to eliminate FEXT interference.
In the prior art, signal processing performed at a central office in the Vectored-DSL is as follows: according to a channel transmission equation y=Hx+n, for a received upstream signal, after the signal processing is performed at the central office, the received signal is y=W(Hx+n)=WHx+Wn, where H is a channel matrix that is used to represent a mutual interference relationship between multiple signals, y is a channel output vector, x is a channel input vector, n is a noise vector, W is a cancellation matrix, and when WH is a diagonal matrix, an FEXT of the received signal is cancelled; for a to-be-sent downstream signal, after the signal processing is performed at the central office, the sent signal is x1=Px, where P is a precoding matrix, so that a received signal obtained by a customer premises equipment is y=Hx1+n=HPx+n, and when HP is a diagonal matrix, the FEXT of the received signal is cancelled. It can be seen that, a key of the signal processing of the Vectored-DSL is to determine the precoding matrix P and the cancellation matrix W, so that FEXT cancellation processing is performed on a signal according to P and W. P and W are determined through estimation during an initialization process of the DSLAM. For example, when a new line is added to the xDSL system, the DSLAM performs the initialization process again to update P and W, and after P and W are determined, activation of the DSLAM is completed and the DSLAM starts to work normally. The foregoing precoding matrix P and the cancellation matrix W are actually formed by multiple corresponding coefficients that are used to represent a mutual interference relationship between lines. For example, the precoding matrix P includes a precoding coefficient that represents interference from line 1 to line 2, and the cancellation matrix W includes a cancellation coefficient that represents the interference from line 1 to line 2. An estimation of P and W actually is an estimation of the precoding matrix and the cancellation matrix thereof. Therefore, the foregoing process of estimating P and W may be referred to as Vector coefficient training, where the Vector coefficient refers to the precoding coefficient and the cancellation coefficient.
FIG. 1 is a schematic flowchart of a method for training a Vector coefficient of a vectored digital subscriber line according to the prior art. As shown in FIG. 1, a new joining line is used as an example. In this case, the central office mainly updates a coefficient related to the joining line, of the existing P and W, such as a precoding coefficient and a cancellation coefficient that represent interference from the joining line (referred to as a joining line) to an existing normal working line (referred to as a showtime line), and a precoding coefficient and a cancellation coefficient that represent interference from another line to the joining line. Current Vector coefficient training includes: the first coefficient estimation performed after a handshake phase of the central office and a customer premises equipment, where the first coefficient estimation includes an OP-VECTOR 1 and an RP-VECTOR1, an OP phase is used to estimate a precoding coefficient, and an RP phase is used to estimate a cancellation coefficient; the second coefficient estimation performed after a channel discovery phase of the central office and the customer premises equipment, where the second coefficient estimation includes an OP-VECTOR 1-1 and an RP-VECTOR1-1; and the third coefficient estimation performed after a training phase of the central office and the customer premises equipment, where the third coefficient estimation includes an OP-VECTOR2, an RP-VECTOR 1-2, an OP-VECTOR2-1, and an RP-VECTOR 2. The first two estimations are performed to estimate a coefficient that represents interference from a joining line to a showtime line twice, and the third estimation is performed to estimate a coefficient that represents interference from another line to the joining line. However, a regular estimation method is adopted for each estimation, such as an orthogonal pilot sequence method, and time required for the orthogonal pilot sequence estimation method multiplies as a system scale increases. For example, when there are 288 lines in a system, time required to activate a central office is about three minutes, which is hard to be accepted by an operator and a user, and therefore, user experience is very poor.